IBM on Friday introduced a new range of systems capable of handling massive amounts of computational data faster at nearly 20 percent better price/performance than comparable Intel Xeon v3 Processor-based systems, according to the company. Designed for processing of massive amounts of data, IBM's new systems leverage the building blocks of the OpenPOWER Foundation, which has Samsung, Google, Tyan and Nvidia as members. The consortium came into existence in 2013 after IBM started licensing its Power architecture so third-party companies can develop chips, servers and components.

The new IBM Power S824L servers are built on IBM’s POWER8 processor, the first processor optimized for the most demanding Big Data workloads. The new systems integrate IBM and other OpenPOWER member technologies, including NVIDIA’s GPU accelerator technology for the first time.

Built on the OpenPOWER stack, the Power S824L systems run data-intensive tasks on the POWER8 processor while offloading other compute-intensive Big Data workloads to GPU accelerators which are capable of running millions of data computations in parallel and are designed to speed up compute-intensive applications.

To take advantage of GPU acceleration on Power Systems, IBM will be optimizing IBM Big Data enterprise applications, including the IBM DB2 database software with BLU Acceleration. Additionally, IBM is working to optimize Power versions of already used GPU-accelerated applications for bioinformatics, defense, finance, molecular dynamics, weather modeling – including SOAP3, NAMD, GROMACS, FFTW library, and Quantum Espresso.

In the future, IBM plans to offer chip- and server-level upgrades for new Power servers that will start shipping in 2016. The servers, which will be faster than Watson and existing systems running on IBM's current Power8 chips, will take advantage of NVIDIA's NVLink technology, which will enable faster data transfers inside servers. NVLink can provide five times more throughput than the PCI-Express pipes used in servers today, Nvidia has said.

Nvidia also has plans to use the NVLink interconnect in its graphics processors, which could then be plugged into IBM's upcoming servers.

IBM will also work with top memory makers to bring new forms of memory into Power servers.

In addition to the GPU-accelerated Power S824L offering, other additions to IBM’s POWER8 processor-based Power Systems portfolio and solutions that complement the lineup include:

• IBM Data Engine for NoSQL: Bringing together technology from OpenPOWER members Redis Labs, Canonical, Altera and IBM, this solution enables a lower cost basis for deploying NoSQL data stores. The solution combines Redis Labs software, Canonical's Ubuntu Linux operating system, Altera’s FPGAs and IBM's CAPI-enabled Power System S822L with IBM's FlashSystem 840. Simplicity is achieved through server consolidation, enabling one POWER8 server to be used instead of 24 Intel-based servers for a well-sized NoSQL store.
• IBM Data Engine for Analytics – Power Systems Edition: Combining POWER8 Scale-out systems with IBM’s flash-based elastic storage technology and Platform Computing software, this new analytics appliance offering can use a third of the storage infrastructure as required by an x86-based solution.
• Power Enterprise Systems: The new Power Enterprise Systems expand the lineup from the POWER8 based scale-out servers and are designed and optimized for the demands of enterprise data and computing environments. The new Power E870 and Power E880 Systems are the highest performance 8-socket systems in the industry with support for up to 1,000 VMs per system. With choices of 32, 40 or 48 processor cores and up to 4 TB per compute node, these systems are designed for modular efficiency and growth and will grow to a massive 192 cores with more than 1500 threads of compute power and 16 TB of memory. They are capable of running running AIX, IBM i, and Linux operating systems.
• Power Enterprise Pools: Enterprise Pools offer flexibility to changes in workload or infrastructure. They enable IBM's clients to move resources across a pool of systems to match workload demand.